内窥镜图像病变特征快速定位算法的仿真分析

摘要

The paper proposed an effective image area mining method based on an improved particle swarm optimization algorithm. The capsule endoscope image features are extracted, and a more simple mapping contact is established. Only the difference values of characteristics and characteristics difference coefficients are calculated. The effective area of the capsule endoscope image is mined, thereby lesion area is quickly extracted from the image. Experimental results show that the improved algorithm is effective.%针对人体内肠道结构复杂,蠕动距离很长,涉及图像复杂性高,存在大量的与定位特征高度类似的冗余干扰特征问题,传统特征高斯定位模型在对肠道海量冗余特征干扰下的定位过程中,由于相似特征众多,会使得算法陷入无限特征对比的境地,阀值很难确定,定位准确性较低.为了解决上述问题,提出了一种改进粒子群优化算法的图像有效区域挖掘方法.提取胶囊内窥镜图像特征,并建立较为简单的映射联系,只计算关系内的特征分量差值和特征差异系数,对胶囊内窥镜图像有效区域进行挖掘处理,从而快速提取出与病变区域相关的图像.实验结果表明,改进算法能够有效提高胶囊内窥镜有效区域定位的效率.